An action or event is biased if it shows a consistent pattern over time. For example: a pair of dice will always produce more eights than twelves, because there are more ways to create an eight with two dices (2+6, 4+4, 5+3) than twelves that can be created only with one pair of 6. But if you change the dice weight by adding more weight on each die, exactly on the side where the sixes are, you will create a bias and the probability of getting 12 will be higher. The added weight creates a bias.
The same biases you will find in lottery games. Although, in theory, each lottery drawing has nothing to do with the previous drawings, in real life this is not true. We cannot create a lottery machine that is perfectly ideal and that obeys all of the principles we have in theory. There will always be a difference between hypothetical theory and real life experiments. This difference is a result of the machine’s imperfect symmetry, the air convection when mixing the balls, the friction between balls, between balls and machine walls, and many more factors. Because of this there will always be some “preferences” for some numbers.
Despite what the lottery commissions claim, randomness can never be totally achieved so there will always be some form of bias; and that is exactly what we are trying to exploit.